Some Comments on Short-Run Earnings Fluctuation Bias

Considerable attention has been given in recent years to studies of the cost of capital and investor preferences for dividends, but empirical analysis of the various theoretical propositions has been complicated by numerous difficulties associated with estimating the parameters in question. In many of the empirical studies, ordinary least squares has been the statistical method applied because, under certain assumptions, ordinary least squares produces unbiased estimates with a minimum of variance. However, the validity of one of the assumptions, namely, that all independent variables contain no measurement error, has been questioned by numerous writers. More specifically, it has been argued that investors tend to ignore temporary or random disturbances in current reported earnings. Thus, values of current earnings that are actually observed for the statistical study may be different from those values which ideally would be observed; i.e., they may contain measurement error, the size of which is reflected by the difference between the current reported earnings and the values on which investors presumably base their decisions. For example, in such least squares regressions as price on dividends and retained earnings, the impact of the measurement error is considered to fall mainly on the retained earnings, since dividends are thought to be relatively stable.